EBV infected cells in the multiple sclerosis brain express PD-L1: How the virus and its niche may escape immune surveillance

The presence of EBV infected B cells in postmortem multiple sclerosis (MS) brain tissue suggests immune evasion strategies. Using immunohistochemical techniques we analysed the expression of the immune checkpoint molecule PD-L1 and its receptor PD-1 in MS brains containing B cell-enriched perivascular infiltrates and meningeal follicles, a major EBV reservoir. PD-1 and PD-L1 immunoreactivities were restricted to CNS-infiltrating immune cells. PD-L1 was expressed on B cells, including EBV infected B cells, while PD-1 was expressed on many CD8 + T cells, including EBV-specific CD8 + T-cells, and fewer CD4 + T cells. PD-L1 + cells and EBV infected cells were in close contact with PD-1 + T cells. PD-L1 expressed by EBV infected B cells could favour local immune evasion leading to EBV persistence and immunopathology in the MS brain.


Introduction
Multiple sclerosis (MS) is the most common chronic inflammatory disease of the central nervous system (CNS), resulting from interactions between multiple predisposing genes and facilitating environmental factors.Among the latter, infection with Epstein-Barr virus (EBV), a common B-lymphotropic herpesvirus, has been identified as an essential factor for MS development (Bjornevik et al., 2022).As recently reviewed (Bjornevik et al., 2023), it is now well established that: i) there is no MS without prior EBV infection; ii) previous infectious mononucleosis and high anti-EBV antibody levels [mainly anti-EBV nuclear antigen 1 (EBNA1) IgG] increase MS risk and independently interact with the main genetic MS risk factor, the HLA-DRB1*1501 allele; iii) EBNA1 IgG titers remain elevated after MS onset; and iv) T cell-mediated immunity toward EBV is qualitatively and quantitatively altered in MS patients.Altogether, these findings suggest that MS is characterized by an early and continuous dysregulation of EBV-specific immune control.
The mechanisms through which EBV infection initiates and possibly sustains CNS inflammation in susceptible individuals are still debated (Aloisi et al., 2023).The hypothesis that EBV might cause CNS autoimmunity through molecular mimicry is supported by the identification of antibodies or T cells that cross-recognize peptides from EBV proteins (mainly EBNA1) and proteins expressed exclusively in the CNS or with a wider distribution (Lünemann et al., 2008;Tengvall et al., 2019;Lanz et al., 2022;Thomas et al., 2023).Other less documented EBV-mediated mechanisms that might lead to CNS autoimmunity include immortalization of B cells producing autoantibodies and inducing autoreactive T cell responses (Pender, 2003), and induction of autoantigens (van Noort et al., 2023).EBV-induced activation of human endogenous retroviruses leading to production of neurotoxic mediators was also proposed (Dolei, 2018).
The presence of EBV infected B cells in the brain of MS patients raises the question as to whether CNS-recruited immune cells might fail to eradicate the infection due to viral evasion mechanisms.We have shown that EBV persists and reactivates in ectopic B cell follicles that form in the meninges of MS patients (Serafini et al., 2004(Serafini et al., , 2007;;Angelini et al., 2013) and are associated with extensive cortical damage and more severe disease progression (Magliozzi et al., 2007(Magliozzi et al., , 2010)).These ectopic lymphoid structures likely represent "immunoprivileged niches" for viral survival and, therefore, major sites where immune evasion might take place.
Both tumors and viruses use several strategies to evade immune surveillance.EBV is one of the most complex viruses with nearly 100 genes encoded by its genome.EBV genes are responsible for clonal proliferation and survival of the infected B cells during latent infection, and for viral particle production leading to new infection events during the lytic phase (Thorley-Lawson, 2015;Damania et al., 2022).Importantly, EBV gene expression is also critical for the virus's ability to evade the host immune response through a wide variety of mechanisms interfering and compromising innate and adaptive immunity.These include downregulation of most EBV proteins and expression of small regulatory RNAs (miRNAs) with immunosuppressive activity during latent infection, and inhibition of molecules involved in antigen processing and presentation, a general reduction of host cell gene expression, and synthesis of an immunosuppressive IL-10-like molecule (BCRF1) during the lytic phase (Ressing et al., 2015;Albanese et al., 2022).EBV-associated tumors use immune evasion mechanisms to promote cell survival, including downregulation of viral proteins, downregulation of major histocompatibility complex (MHC) molecules, and induction of the immunosuppressive enzyme indoleamine 2, 3-dioxygenase and of the inhibitory immune checkpoint programmed deathligand 1 (PD-L1) (Farrell, 2019).
PD-L1 is expressed or inducible in different cell types, including antigen-presenting cells, lymphocytes and stromal cells; it engages its receptor PD-1 on the T cell surface to inhibit T cell activation (proliferation, cytokine production, cytolytic function) (Riella et al., 2012;Pauken et al., 2021).PD-1 is a co-receptor of the CD28/CTLA-4 family, expressed on lymphocytes previously activated by interaction with the corresponding antigen (Riella et al., 2012).In physiological conditions, the PD-1/PD-L1 interaction is crucial for regulating immune responses and has also been implicated in the germinal center reaction, since PD-L1 on follicular dendritic cells (FDC) and stromal cells interacts with PD-1 on follicular T helper cells to influence B cell survival and clonal expansion (Heesters et al., 2021).The PD-1/PD-L1 axis can also promote the expansion of regulatory T cells, which further hampers T cell effector functions.In chronic viral infections and malignancies, PD-L1 plays a key role in the progressive loss of T cell and natural killer (NK) cell effector functions, resulting in immune exhaustion (McLane et al., 2019;Merino et al., 2020).
PD-L1 is highly expressed in EBV-associated tumors, like B cell lymphoma and naso-pharyngeal carcinoma, and is upregulated by the EBV latent proteins EBNA2 (Anastasiadou et al., 2019;Yanagi et al., 2021) and latent membrane protein-1 (LMP1) (Fang et al., 2014;Bi et al., 2016), and by EBV reactivation (Yanagi et al., 2022), to inhibit the anti-tumoral immune response.PD-L1+ EBV+ B cells have been also detected in the tonsils of patients with infectious mononucleosis, suggesting a role for the PD-1/PD-L1 axis in modulating the immune response to EBV during acute infection (Barros et al., 2019).
Aiming at better defining the mechanisms contributing to EBV persistence in the CNS of MS patients, we have performed an immunohistochemical study of PD-L1 and its receptor PD-1 in postmortem brain samples from cases with progressive MS and non neuropathological control cases.We have investigated the expression of PD-L1 in B cells, EBV infected B cells, myeloid cells and FDC in white matter lesions and meninges of MS cases, and have asked whether T cells, including EBVspecific CD8+ T cells, express PD-1 and interact locally with PD-L1expressing cells.

Tissue samples
Human postmortem brain tissues from six cases with progressive MS and four control cases without neurological disease were analysed in this study (Supplementary Table 1).All brain samples from MS and nonneurological control cases were provided by the UK Multiple Sclerosis Tissue Bank at Imperial College London (https://www.imperial.ac.uk /medicine/multiple-sclerosis-and-parkinsons-tissue-bank/).Based on the available clinical histories, all MS cases were in the progressive phase of the disease and were wheelchair-or bed-bound at the time of death (Expanded Disability Status Scale >7).No treatment was reported during the progressive MS phase.For MS cases, mean disease duration was 20.3 ± 9 years and mean age at death 45.2 ± 7 years.Donor information is provided in Supplementary Table 1.
Control lymphoid tissues included one autoptic abdominal lymph node from a control subject, two tonsils from two children with recurrent bacterial infections undergoing tonsillectomy (provided by the Institute of Pathological Anatomy, U.C.S.C. Policlinico A. Gemelli, Rome, Italy), and two tonsils from two cases of EBV-associated infectious mononucleosis (kindly provided by Prof. Gerard Niedobitek, Institute of Pathology, Sana Klinikum Lichtenberg, Berlin, Germany).Use of human tissue for research purposes was approved by the Ethics Committee of the Istituto Superiore di Sanità (CE 12/356).

Neuropathological assessment
For each MS case, well preserved cerebral tissue blocks with substantial B cell infiltration in the white matter (WM) and meninges and presence of intrameningeal B cell follicles were selected.Twelve tissue blocks (4 cm 3 ) from superior frontal gyrus, precentral gyrus and middle temporal gyrus from six MS cases were analysed (Supplementary Table 1).Seven brain blocks from five MS cases were fixed in 4% paraformaldehyde (PFA) and frozen (FF samples) and stored at − 80 • C until use.Four brain blocks from four MS cases were snap frozen (SF samples), and one brain block was formalin-fixed and paraffin embedded (FFPE sample).All brain blocks from control cases without neurological disease and the abdominal lymph node were FF; three tonsils were FFPE and one tonsil was SF.
Series of 10-μm and 4-μm thick sections were cut from the selected FF or SF and FFPE tissue blocks, respectively.Neuropathological features of MS tissues (extent of demyelination, lesion inflammatory activity, B. Serafini et al. degree of immune cell infiltration in WM and meninges) and lymphocyte composition of immune infiltrates were assessed on the first sections of each series by histological and immunoistochemical stainings, as previously described (Serafini et al., 2019(Serafini et al., , 2023)), and WM lesions were classified according to Kuhlmann et al., 2017.Seven reactive normal appearing WM (NAWM) areas with activated MHC class II+ microglia and a variable degree of perivascular immune cell infiltration, four active and seven chronic active lesions were identified.Substantial B and T cell infiltration was detected in the meninges and WM of all MS tissue blocks selected; eight meningeal B cell follicles were identified in seven blocks from five MS cases.Negligible meningeal and parenchymal immune infiltration was detected in control brains.Subsequent serial sections of each series were analysed using in situ hybridization (ISH) for EBV-encoded small RNA (EBER) and immunohistochemistry for the selected viral and cell markers.
Tonsils and lymph node were stained with hematoxylin and eosin for histological evaluation, and analysed using EBER ISH (FF and FFPE tissues) and immunohistochemistry for the selected viral and cellular markers.

Immunohistochemistry
SF and FF sections were air dried for two hours at room temperature (RT) and then fixed in 4% PFA for 10 min at RT or in cold acetone for 10 min at 4 • C, respectively.SF sections were then rinsed twice in distilled water and then in phosphate buffered saline (PBS); FF sections were air dried for 10 min at RT. FFPE sections were dewaxed in xylene and rehydrated in decreasing series of ethanol.After rehydration in PBS, FF and FFPE sections were subjected to the appropriate antigen retrieval treatment.Antibodies, staining conditions and antigen retrieval treatments are shown in Supplementary Table 2.
For double stainings in bright field, the Zytochem plus 2-step double stain polymer kit (Zytomed Systems GmbH, Berlin, Germany) was used, as previously described (Serafini et al., 2023).Briefly, after incubation with a combination of rabbit and mouse primary antibodies, sections were treated sequentially with AP-Polymer anti-rabbit, Tris buffered saline (TBS), permanent AP Red Kit containing Levamisol (Vector Lab), TBS, horseradish peroxidase (HRP)-conjugated anti-mouse polymer, and permanent HRP Green Kit (green/blue colour; Zytomed System).
For CD20 and PD-1 double staining, sections were first incubated with anti-CD20 mAb, biotinylated rabbit anti-mouse Ig, ABC and DAB solution, and then sequentially with anti-PD-1 mAb, HRP-conjugated anti-mouse polymer and permanent HRP Green Kit.After very quick washing in distilled water, sections were sealed with Ultramount Aqueous permanent mounting medium (Agilent Dako, Santa Clara, CA, USA) and analysed with an Axioscope microscope (Carl Zeiss, Jena, Germany) equipped with AxioCam MRc5 camera and Axiovision SE 64 software.
Double immunofluorescence stainings for CD20 and LMP2A or PD-L1, LMP1 and LMP2A, PD-L1 or PD-1, were performed using the appropriate combination of the fluorophore-conjugated secondary antibodies (2 μg/ml; Thermo Fisher Scientific, Waltham, MA, USA, and Jackson Laboratories), as described (Angelini et al., 2013).Sections were sealed with Prolong Gold antifade reagent with DAPI (Invitrogen, Thermo Fisher Scientific).Sections were analysed with ZEN 3 Lite software and images were acquired with an epiluminescence microscope (Zeiss Axioscope) equipped with an Axiocam 512 digital camera.
After fixation with 2% PFA and washing in PBS, sections were incubated ON with a mixture of rabbit anti-R-PE polyclonal antibody (GeneTex Inc., Irvine, CA) and anti-PD-1 mouse mAb in PBS containing 1% bovine serum albumin.After washing in PBS, sections were incubated with a mixture of biotinylated goat anti-rabbit IgG and Alexa Fluor 488-conjugated donkey anti-mouse IgG (Invitrogen) diluted 1:150 in PBS containing 1% normal donkey serum and then with tetramethylrhodamineconjugated streptavidin (Jackson ImmunoResearch Laboratories).After further washing in PBS, sections were mounted with antifade mounting medium containing DAPI.Sections were analysed and images acquired with a digital epifluorescence microscope (Axioscope Zeiss).Negativecontrol stainings were performed using mismatched pentamers and Ig isotype controls.

Cell counts and localization in MS brain sections
EBER+ EBV-infected cells and CD20+ cells were counted in serial brain sections from four MS cases; the fraction of EBV infected B cells was calculated indirectly, by dividing the number of EBER+ cells by the number of CD20+ cells counted in B-cell enriched perivascular infiltrates in the WM (n = 4) and meningeal B cell follicles (n = 5).CD20+ B cells, EBV infected LMP2A+ cells, PD-L1+ cells, and double positive PD-L1+ CD20+ and PD-L1+ LMP2A+ cells were counted in eight brain sections double-stained for CD20 and PD-L1, and eight brain sections double-stained for LMP2A and PD-L1 from eight tissue blocks of five MS cases.The percentages of cells co-expressing PD-L1 and CD20 or LMP2A were calculated relatively to the total CD20+, LMP2A+ and PD-L1+ cell populations in WM perivascular infiltrates of different size (n = 9) and B cell follicles (n = 6).The percentages of cells co-expressing PD-1 and CD4 or CD8 in the total CD4+ and CD8+ T cell populations were calculated in six serial sections from three brain tissue blocks of three MS cases.All cell counts were performed manually, with a morphometric grid and 20× and 40× objectives.Cell percentages obtained in different brain tissue blocks and median values are shown in Supplementary Fig, 2; in the text, data are presented as median values and ranges.

Expression of PD-L1, PD-1 and EBV RNA in lymphoid tissues
Preliminary experiments were performed in an autoptic, non pathological abdominal lymph node from a control subject and in tonsillectomy samples from two children with recurrent bacterial infections and two individuals with EBV-associated infectious mononucleosis, aiming at establishing the appropriate immunohistochemical procedures for PD-L1 and PD-1 detection in tissues processed in different ways (SF, FF, FFPE) and substantiate expression of PD-L1 in EBV infected B cells (Barros et al., 2019).
In the FF abdominal lymph node, PD-L1 expression was mainly localized in cells showing a stromal/dendritic cell morphology in the interfollicular T cell area (Fig. 1A).Numerous PD-1+ lymphocyte-like cells were detected in the T cell area and in the germinal center of B cell follicles, likely follicular T helper cells (Fig. 1B).No EBV infected cells were detected using EBER ISH (data not shown).In the tonsils from two children with recurrent bacterial infections (one SF and one FFPE, respectively), PD-L1 expression was confined to the epithelial cells of the tonsillar crypts (Fig. 1 C), and PD-1 was detected in lymphocytes in the B cell follicles and in the interfollicular T cell area (Fig. 1 D).EBER+ cells were almost absent; occasionally, one isolated EBER+ cell was found inside a follicle (Fig. 1E).In the infectious mononucleosis tonsils (both FFPE), PD-L1 was expressed in numerous perifollicular lymphocytes, in addition to the epithelial cells of the tonsillar crypts (Fig. 1 F, G).As shown previously (Croia et al., 2013), EBER+ cells were present at high density in the T cell area but were not or only rarely observed inside the B cell follicles (Fig. 1 H).PD-L1 was detected in some of the perifollicular cells expressing nuclear EBER (Fig. 1 I) or the EBV latent protein LMP2A (Fig. 1 J).PD-1 was expressed in lymphocytes in the perifollicular area and the germinal center of B cell follicles (Fig. 1 K, L).Non pathological abdominal lymph node (A, B): immunohistochemical stainings for PD-L1 and PD-1 reveal the presence of PD-L1+ cells in the interfollicular T cell area (A and inset) and PD-1+ lymphocyte-like cells in the interfollicular T cell area and in the germinal center (GC) and at the periphery of a B cell follicle (B).Palatine tonsil from a child with recurrent bacterial infections (C-E): immunostaining for PD-L1 reveals expression of PD-L1 in the reticular epithelium of the tonsillar crypts (C, arrow); the inset in C shows the membrane localization of PD-L1 in the crypt epithelial cells at high power magnification.Immunostaining for PD-1 shows lymphocyte-like PD-1+ cells in the germinal center (GC) of B follicles and in the interfollicular area (D).EBER ISH (E; nuclear blue staining) shows virtual absence of EBV infected cells (one EBER+ nucleus is indicated by the arrow in E).Tonsil from a case with EBV-associated infectious mononucleosis (F-L): Immunohistochemical staining for PD-L1 (F, G) shows PD-L1 immunoreactivity in the crypt epithelial cells (F, arrow) and numerous lymphocytic cells in the T cell area (G).EBER ISH shows that EBER+ nuclei localize mainly in the T cell area and are less frequent in the B cell follicles (H).EBER ISH (blue nuclei) combined with PD-L1 immunohistochemistry (red/staining) reveals that some EBER+ cells co-express PD-L1 (I); the inset in I shows the nuclear localization of EBER and the membrane localization of PD-L1 in the cell indicated by the arrow in I at high power magnification.Double immunohistochemistry for PD-L1 (red) and the EBV latent membrane protein LMP2A (blue) shows that PD-L1 is expressed on most LMP2A+ cells (J).The bottom insets in J show the double positive PD-L1+ LMP2A+ cells indicated by the arrows at high power magnification; in the top inset in J, double immunofluorescence staining for CD20 (green) and LMP2A (red) shows expression of LMP2A in CD20+ cells.Immunostaining for PD-1 in the infectious mononucleosis tonsil shows PD-1+ cells in the germinal center (GC) (K,L) and at the periphery of a B cell

Expression of PD-L1 in postmortem control and MS brain tissues
No CD20+ B cells and EBER+ cells were detected in any of the control non neuropathological brains analysed (n = 4).In the same samples PD-L1+ cells were not observed (Supplementary Fig. 1 A).As previously described (Serafini et al., 2004(Serafini et al., , 2007(Serafini et al., , 2010(Serafini et al., , 2019;;Angelini et al., 2013), in the WM of cases dying during the progressive phase of MS CD20+ B cells were mainly found in the perivascular space and their frequency was highly variable in and between WM lesions.B cell enriched perivascular cuffs were typically, but not exclusively, observed in actively demyelinating lesions (Fig. 2 A).In the meninges, CD20+ B cells were scattered in the diffuse infiltrates and were tightly packed inside B cell follicle-like structures forming in the subarachnoid space of the most infiltrated MS brains (Fig. 2 C).
We have shown previously that a substantial proportion of MS braininfiltrating B cells are EBV infected (Serafini et al., 2007(Serafini et al., , 2010)).In the MS brain samples analysed in this study, the distribution and frequency of EBV infected cells, identified using EBER ISH (Fig. 2 B, D) and double immunofluorescence stainings for the EBV latent proteins LMP1 and LMP2A (Fig. 2 E-G), roughly follow the same pattern.Specifically, EBER+ cells accounted for 35% (median value, range 24-45) and 58% (median value, range 30-72) of CD20+ B cells in WM perivascular immune infiltrates and meningeal follicles, respectively.Data obtained in each MS case analysed are shown in Supplementary Fig. 2.
To assess whether CNS-infiltrating EBV infected B cells express PD-L1, we first stained consecutive MS brain sections for CD20 and PD-L1.We observed that PD-L1+ cells almost exclusively localized in B cell-enriched perivascular cuffs in the WM (Fig. 3 A, B; Supplementary Fig. 3) and in the B cell core of meningeal follicles (Fig. 3 C-F).Conversely, no or rare PD-L1+ cells were detected in perivascular immune infiltrates devoid of B cells in the MS WM (Supplementary Fig. 3) and meninges.Double immunofluorescence staining for CD20 and PD-L1 confirmed expression of PD-L1 in CNS-infiltrating B cells (Fig. 3 G-I).PD-L1 was expressed in 36% (median value, range 19-75) and 17% (median value, range 15-65%) of CD20+ cells in B cell-enriched WM perivascular infiltrates and meningeal B cell follicles, respectively.PD-L1+ CD20+ cells accounted for 87% (median value, range 78-90) and 43% (median value, range 34-56) of total PD-L1+ cells in B cellenriched WM perivascular infiltrates and meningeal B cell follicles, respectively.Data obtained in each MS case analysed are shown in Supplementary Fig. 2.
We then double stained MS brain sections for EBER and PD-L1.However, the combination of EBER ISH and PD-L1 immunostaining performed in brain sections from one FFPE sample and three FF samples did not result in an appreciable staining for PD-L1, likely due to the combination of fixation conditions and processing for EBER ISH of autoptic brain tissue.We therefore performed double immunofluorescence staining for PD-L1 and the EBV latent protein LMP2A and found that a large proportion of LMP2A+ cells in WM perivascular cuffs (61% median value, range 23-100) (Fig. 4) and B cell follicles (86% median value, range 55-100) (Fig. 5 A-N) co-expressed PD-L1.LMP2A+ PD-L1+ cells accounted for 56% (median value, range 42-69%) and 45% (median value, range 33-66%) of total PD-L1+ cells in B cell enriched perivascular WM infiltrates and meningeal B cell follicles, respectively.Data obtained in each MS case analysed are shown in Supplementary Fig. 2.
Since FDC in lymphoid tissue express PD-L1 (Heesters et al., 2021), we performed double stainings for the FDC marker CD35 and PD-L1 in MS brain sections containing meningeal B cell follicles.CD35+ PD-L1+ cells accounted for about 15% of the PD-L1+ cell population observed in two B cell follicles, whereas PD-L1+ cells with a lymphocyte-like morphology were CD35 negative (Fig. 5 O, P).No PD-L1+ CD35+ cells where observed in the diffuse meningeal infiltrates (data not shown).
To assess whether PD-L1 is expressed on other antigen presenting cells in the MS brain we performed double stainings in bright field and immunofluorescence for PD-L1 and CD68, a membrane glycoprotein expressed by myeloid lineage cells, mainly monocytes and macrophages.Some perivascular CD68+ macrophages in the reactive NAWM and WM lesions, and very few CD68+ macrophages/amoeboid microglia in actively demyelinating WM lesions were found to co-express PD-L1 (Supplementary Fig. 4 A-E).Reactive ramified microglia was largely PD- L1 negative; only occasionally, PD-L1+ CD68+ microglial cells were detected in the reactive NAWM (Supplementary Fig. 4 F).To verify whether PD-L1 is expressed on astrocytes, we also performed double stainings for PD-L1 and GFAP, both in bright field and immunofluorescence.We observed that GFAP+ astrocytes in the MS WM do not express PD-L1; PD-L1+ GFAP+ astrocytes were rarely detected only in the grey matter (Supplementary Fig. 4 G-K).This study does not confirm the results of previous studies showing PD-L1 immunoreactivity predominantly on astrocytes in experimental autoimmune encephalomyelitis and postmortem MS brain samples (Linnerbauer et al., 2023;Smith et al., 2023).

MS brain-infiltrating T cells express PD-1 and contact B cells and EBV infected cells
In non-neurological control brains (n = 4), PD-1 immunoreactivity was detected in rare, isolated cells with a lymphocyte morphology in two cases (Supplementary Fig. 1 B, C).We next evaluated PD-1 expression on CD4 and CD8 T cells and the relationship of PD-1+ cells to PD-L1+ cells, CD20+ B cells and EBV infected cells in the MS brain.Immunostainings performed in serial sections showed a similar distribution of CD8+ cells and PD-1+ cells (Supplementary Fig. 3).Double immunostaining for PD-1 and CD4 showed that PD-1+ CD4+ cells were consistently observed in the inflammatory infiltrates, accounting for 10% (median value, range 4.4-15) of total CD4+ cells in WM perivascular immune infiltrates and meninges (Fig. 6 A, B; Supplementary Fig. 2).Double immunostaining for PD-1 and CD8 confirmed expression of PD-1 on a substantial proportion of CD8+ cells in WM lesions and meninges (Fig. 6 C-E).PD-1+ CD8+ cells accounted for 42% (median value, range 26-77) of the total CD8 T cell population in in WM perivascular infiltrates and meninges (Supplementary Fig. 2).
Double stainings in bright field and immunofluorescence for Iba-1 and PD-1 showed that most ramified microglia and macrophages/ amoeboid microglia in the MS WM parenchyma do not express PD-1 (Supplementary Fig. 4 L, M).Only occasionally, isolated PD-1+ Iba-1+ microglial cells were detected in the reactive NAWM (Supplementary Fig. 4 N).Similarly to PD-L1, this study does not confirm the results of previous studies showing predominant expression of PD-1 on microglial cells, and absence or negligible expression of this molecule on tissue-infiltrating lymphocytes in the MS brain (Pittet et al., 2011;Linnerbauer et al., 2023;Smith et al., 2023).
In two MS cases carrying the HLA-B*0801 allele (MS92 and MS121; Supplementary Table 1) it was possible to stain fresh frozen brain sections with B*0801 pentamers coupled with peptides from the EBV latent protein EBNA3A and the EBV lytic protein BZLF1 to identify EBVspecific CD8+ T cells and assess if they expressed PD-1.Staining of brain sections with anti-PD-1 antibody and pooled B*0801/EBNA3A and B*0801/BZLF1 pentamers confirmed the presence of EBV-specific CD8+ T cells in the inflammatory infiltrates of both MS cases (Serafini et al., 2019) and revealed that virtually all EBNA3A/BZLF1 pentamers+ CD8+ T cells expressed PD-1 (Fig. 7).
Finally, double immunostainings of MS brain sections allowed to visualize direct contacts of PD-1+ cells with PD-L1+ cells (Fig. 8 A-C), CD20+ cells (Fig. 8 D -F), and LMP2A+ cells (Fig. 8 G-J) in WM perivascular immune infiltrates and in the meninges, mainly at the periphery of and outside the B cell follicles.

Discussion
In this study we have asked whether the PD-1/PD-L1 pathway might contribute to immune evasion by EBV infected B cells that accumulate in the MS brain, and presumably promote the recruitment of virus-specific memory and/or effector CD8+ T lymphocytes, while inhibiting their local reactivation.The main finding is that PD-L1 is expressed in a fraction of B cells and most EBV latently infected B cells present in WM lesion perivascular cuffs and in the inflamed meninges, particularly in ectopic B cell follicles.Accordingly, PD-L1+ EBV infected B cells were detected in the tonsils of individuals with acute EBV infection (Barros et al., 2019; this study), but not recurrent bacterial infections.Intriguingly, in the infectious mononucleosis tonsil EBV infected B cells and PD-L1+ cells localize predominantly outside the B cell follicles, while the opposite was found in the MS brain, which argues that EBV infection has a leading role in PD-L1 induction.This observation is consistent with the results of previous studies showing that EBV-encoded EBNA-2 (Anastasiadou et al., 2019;Yanagi et al., 2021), LMP1 (Fang et al., 2014;Bi et al., 2016) and miRNAs (Rang et al., 2022) induce PD-L1 on EBV infected B cells.Furthermore, EBV products can be packaged into extracellular vesicles that are released from EBV infected cells and affect the behaviour of neighbouring and distant cells (Cone et al., 2019).LMP1 and miRNAs are present in extracellular vesicles derived from EBV infected cells (Vazirabadi et al., 2003;Cone et al., 2019) and LMP1 increases the amount of PD-L1 in these particles (Abou Harb et al., 2023), suggesting a possible contribution to the immunosuppressive microenvironment.Besides EBV, pro-inflammatory cytokines, particularly interferons, could also contribute to induction of PD-L1 expression in B cells (Sun et al., 2018;Schönrich and Raftery, 2019).
PD-L1+ EBV infected B cells were found in brain sections from all the In previous studies we observed that some CNS-infiltrating CD8+ T cells are cytotoxic, namely express granzyme B and the degranulation marker CD107a, and contact EBV infected B cells, mainly in active WM lesions and at the periphery of meningeal B cell follicles (Serafini et al., 2007(Serafini et al., , 2013)).We also showed that CNS-infiltrating, EBV-specific CD8+ T cells express CD107a and contact CD20+ B cells or LMP2A+ EBV infected cells (Serafini et al., 2019).Here, we show that all EBV-specific CD8+ T cells express PD-1 and that PD-1+ cells contact PD-L1+ cells, B cells as well as LMP2A+ EBV infected cells.Altogether, these findings suggest that the continuous engagement of CNS-recruited, EBV-specific CD8+ T cells induces the production of cytotoxic factors, but fails to eradicate EBV.Such ineffective antiviral immune response likely causes secondary neural tissue damage in MS.
Persistent antigenic stimulation may lead to the exhaustion of CD8+ T cells and NK cells, both of which have an essential role in the elimination of virus infected cells (McLane et al., 2019;Merino et al., 2020).Accordingly, several studies have demonstrated a progressive decrease in the frequency of EBV-specific CD4+ and CD8+ T cells with increasing MS disease duration, and reduced cytokine polyfunctionality and cytotoxicity of EBV-specific CD8+ T cells in the peripheral blood of MS patients (Pender et al., 2017;Jilek et al., 2012;Cencioni et al., 2017).Expansion of circulating EBV-specific CD8+ T cells during active disease (Angelini et al., 2013;Pender et al., 2017), and broadened specificity of EBV latent and lytic antigen-specific T cells in MS patients compared to healthy seropositive subjects suggest increased viral antigen presentation in MS (Lünemann et al., 2006;Schneider-Hohendorf et al., 2022).The predominant expression of PD-1 on CNS-infiltrating T cells documented in this study is consistent with studies showing expression of PD-1 on CD8+ T cells, including tissue resident memory T cells, in brain lesions and in the CSF of patients with MS (van Nierop et al., 2017;Smolders et al., 2018;Fransen et al., 2020;Serafini et al., 2023).A recent study reported that PD-1+ proinflammatory helper and cytotoxic T cells and PD-L1+ antigen presenting cells, including B cells, are increased in the peripheral blood of patients with relapsing remitting MS and that the levels of soluble and cell-associated PD-L1 correlate with disease disability and MRI activity (Tsaktanis et al., 2023).Altogether, these results are compatible with T cell exhaustion driven by chronic antigenic stimulation.
Decreased frequency and impaired maturation and function of NK cells were found in MS (Beliën et al., 2022), while enhanced NK cell function has been associated with clinical improvement and favourable therapy outcome (Laroni, 2019;McKinney et al., 2021).The NK cell response to EBV is preferentially targeted to the lytic phase of infection and involves activating and inhibitory receptors (Münz, 2022).PD-1 surface expression and inhibitory function have been described also in NK cells (Quatrini et al., 2020) and PD-1 upregulation on NK cells has been reported in several EBV-associated diseases (Desimio et al., 2023).So far, no changes in PD-1 expression on circulating NK cells have been observed in patients with relapsing remitting MS (Tsaktanis et al., 2023).In our preliminary experiments in postmortem MS brain samples, no putative NK cells, defined as NKG2D+ or CD56+ CD3-cells, were detected in WM lesions from three MS cases; isolated CD56+ CD3-were occasionally observed only in the MS meninges (unpublished data).While these results did not allow to examine PD-1 expression on CNSinfiltrating NK cells, NK cell presence in MS WM lesions and meninges cannot be excluded.Because NK cells responding to EBV lytic infection are mainly CD56 dim and chronic engagement of NKG2D results in its downregulation (Münz, 2022), these markers could be weakly expressed on CNS-infiltrating NK cells and not detectable using conventional immunohistochemical techniques.Further studies are required to assess the presence and phenotype of NK cells in the MS brain.Since the EBV life cycle and the activation and regulation of anti-EBV immunity mainly occurs in lymphoid tissue, it is likely that CNS-draining, deep cervical lymph nodes are the main site where chronic EBV dysregulation leading to immune exhaustion occurs (Serafini et al., 2014).Verification of this hypothesis requires further investigations, but limited access to patient lymphoid tissue is a major obstacle.
The strength of this study relies on the analysis of well characterized brain tissue blocks containing prominent immune infiltrates in WM lesions and meninges, and the use of double immunostainings for the simultaneous detection of immune checkpoint molecules and markers of EBV infection or anti-EBV immunity in CNS-infiltrating immune cells.However, several study limitations must be considered.The number of brain samples analysed is limited and all samples are from progressive MS cases.Despite these limitations, our study strongly suggests that the PD-1-PD-L1 pathway might be implicated in EBV persistence in the MS brain.This finding reinforces the concept that improving immune control of EBV infection might ameliorate MS.Blockade of immune checkpoints, including PD-1 or PD-L1, is used to enhance immune responses against tumors.However, onset of MS and increases in reported relapses in patients with MS receiving checkpoint inhibitors indicate that these drugs might exacerbate MS-associated immunopathology (Manenti et al., 2022).To assess whether improved control of EBV infection has a beneficial effect in MS, existing or newly developed antivirals (Drosu et al., 2021;Monaco et al., 2023), potentiation of antiviral immunity with EBV-specific adoptive immunotherapy (Ioannides et al., 2021), and a therapeutic vaccine (Bjornevik et al., 2023) are emerging as potentially promising approaches.

Declaration of competing interest
The Authors have no competing interest to declare.
B.Serafini et al.   (caption on next page) B.Serafini et al.

Fig. 1 .
Fig.1.Localization of PD-L1, PD-1 and EBV RNA in lymphoid tissues.Non pathological abdominal lymph node (A, B): immunohistochemical stainings for PD-L1 and PD-1 reveal the presence of PD-L1+ cells in the interfollicular T cell area (A and inset) and PD-1+ lymphocyte-like cells in the interfollicular T cell area and in the germinal center (GC) and at the periphery of a B cell follicle (B).Palatine tonsil from a child with recurrent bacterial infections (C-E): immunostaining for PD-L1 reveals expression of PD-L1 in the reticular epithelium of the tonsillar crypts (C, arrow); the inset in C shows the membrane localization of PD-L1 in the crypt epithelial cells at high power magnification.Immunostaining for PD-1 shows lymphocyte-like PD-1+ cells in the germinal center (GC) of B follicles and in the interfollicular area (D).EBER ISH (E; nuclear blue staining) shows virtual absence of EBV infected cells (one EBER+ nucleus is indicated by the arrow in E).Tonsil from a case with EBV-associated infectious mononucleosis (F-L): Immunohistochemical staining for PD-L1 (F, G) shows PD-L1 immunoreactivity in the crypt epithelial cells (F, arrow) and numerous lymphocytic cells in the T cell area (G).EBER ISH shows that EBER+ nuclei localize mainly in the T cell area and are less frequent in the B cell follicles (H).EBER ISH (blue nuclei) combined with PD-L1 immunohistochemistry (red/staining) reveals that some EBER+ cells co-express PD-L1 (I); the inset in I shows the nuclear localization of EBER and the membrane localization of PD-L1 in the cell indicated by the arrow in I at high power magnification.Double immunohistochemistry for PD-L1 (red) and the EBV latent membrane protein LMP2A (blue) shows that PD-L1 is expressed on most LMP2A+ cells (J).The bottom insets in J show the double positive PD-L1+ LMP2A+ cells indicated by the arrows at high power magnification; in the top inset in J, double immunofluorescence staining for CD20 (green) and LMP2A (red) shows expression of LMP2A in CD20+ cells.Immunostaining for PD-1 in the infectious mononucleosis tonsil shows PD-1+ cells in the germinal center (GC) (K,L) and at the periphery of a B cell Fig. 2. EBV infected cells in the inflammatory infiltrates in the MS brain.Immunostaining for CD20 (A, C) and EBER ISH (B, D) performed in serial brain sections show the presence of numerous EBV infected cells in a B cell enriched perivascular infiltrate in an active WM lesion (A, B) and in a meningeal B-cell follicle (C, D).Double immunofluorescence for LMP1 (green, E) and LMP2A (red, F) reveals co-expression of these EBV latent membrane proteins (orange, G) in the same meningeal follicle shown in panels C, D. Panels C-G are from serial sections.Nuclei were stained with Meyer's hematoxylin in A and C, and with DAPI in E-G.Bars: 100 μm in A; 50 μm in B-D; 20 μm in E-G.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 3 .
Fig. 3. Expression of PD-L1 in B cell enriched immune infiltrates in the MS brain.Immunohistochemistry for CD20 (A) and PD-L1 (B) performed in serial brain sections shows presence of numerous PD-L1+ cells in a B cell enriched perivascular infiltrate in an active WM lesion.Immunofluorescence staining (C) and immunohistochemistry (E) for CD20, and immunostaining for PD-L1 (D, F) performed in serial sections show presence of numerous PD-L1+ cells in meningeal B cell follicles from two different MS cases (MS 234 in C, D; MS180 in E, F).The inset in F shows the negative control with pre-immune rabbit serum.Double immunofluorescence staining for CD20 (green) and PD-L1 (red) (G-I) shows double labeled PD-L1+ CD20+ cells (orange) in the B cell follicle shown in E and F. Two PD-L1+ CD20+ cells indicated by the arrows in G are shown at high power magnification in H and I; two PD-L1+ CD20-cells are indicated by the arrowheads in G. Nuclei were stained with Meyer's hematoxylin in A, B, E and F, and with DAPI in C, G-I.Bars: 50 μm in A-F; 20 μm in G; 10 μm in H and I. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 4 .
Fig. 4. PD-L1 expression in EBV infected cells in B cell enriched perivascular immune infiltrates.Immunohistochemical staining for CD20 (A) and PD-L1 (B), and double immunofluorescence staining for LMP2A (green) and PD-L1 (red) (C) were performed in serial brain sections from the same MS case (MS234).Double positive PD-L1+ LMP2A+ cells (orange) are detected in a B cell enriched perivascular infiltrate in an active WM lesion (C).Panels D -F show the area comprised in the frame in C at higher magnification.Besides double positive PD-L1+ LMP2A+ cells (arrows in F), PD-L1-LMP2A+ and PD-L1+ LMP2A-cells (arrowheads in F) were also detected.Nuclei were stained with Meyer's hematoxylin in A and B, and with DAPI in C-F.Bars:100 μm in A; 50 μm in B; 20 μm in C-D.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5 .
Fig.5.PD-L1 expression in EBV infected cells in meningeal B cell follicles.Double immunofluorescence staining for LMP2A and PD-L1 was performed in brain sections containing B cell follicles from four tissue blocks of three MS cases.A-C (MS92): the B cell follicle visualized with CD20 immunofluorescence staining (inset in A) contains LMP2A+ cells most of which co-express PD-L1+ (orange, arrows in C and inset; the inset shows a double positive PD-L1+ LMP2A+ cell at higher magnification); PD-L1+ LMP2A-cells are also shown (red, arrowheads in C).D-G (MS92): immunostaining for CD20 (D) and EBER ISH (inset in D) performed in consecutive sections show a B cell follicle containing numerous EBV infected cells; widespread co-expression of LMP2A and PD-L1 is found in this follicle (arrows in G and insets; the insets show several double positive PD-L1+ LMP2A+ cells at higher magnification); PD-L1+ LMP2A-cells are also shown (red, arrowheads in G).H -K (MS330): the B cell follicle visualized with CD20 immunostaining (H) contains LMP2A+ cells most of which co-express PD-L1 (orange, arrows in K); in the same area some PD-L1+ LMP2A-cells (green) are also present (arrowheads in K).L-N (MS180): several double positive PD-L1+ LMP2A+ cells (orange, arrows in N) and two PD-L1-LMP2A+ cells (green, arrowheads in N) are present in the same B cell follicle shown in panel E of Fig.3.Double immunofluorescence staining for CD20 (green) and CD8 (red) (O) shows a B cell follicle surrounded by a few CD8+ cells in a different MS case (MS234); double immunostaining in bright field for PD-L1 (red) and CD35 (blue) performed in a serial section shows that, in the follicle area defined by the frame in panel O, PD-L1 is expressed in some CD35+ stromal cells/FDC (arrows in P; the same cells are shown at higher magnification in the inset).Nuclei were stained with DAPI in A-C, E-G, I -O, and with Meyer's hematoxylin in D and H. Bars: 100 μm in the insets in A and C; 50 μm in D, H, O, P; 20 μm in A-C, E-G, I-N; 10 μm in the insets in C, G and P. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 7 .
Fig. 7. PD-1 expression in MS brain-infiltrating EBV-specific CD8+ T cells.Double immunofluorescence staining with anti-PD-1 mAb (green; A, D) and pooled B*0801/EBNA3A and B*0801/BZLF1 pentamers (red; B, E) performed in brain sections from one HLA-B*0801+ MS donor (MS234) are shown.Merge of PD-1 and pentamer immunostainings (C, F) reveals that PD-1 is expressed on pentamerbinding cells in two perivascular inflammatory infiltrates in an active WM lesion.Arrows in C and F indicate PD-1+ pentamer binding cells (orange), while arrowheads indicate PD-1+ pentamer-negative cells.Nuclei were stained with DAPI.Bars: 10 μm.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 8 .
Fig. 8. Close contacts of PD-1+ cells with PD-L1+ cells, B cells and EBV infected cells in the MS brain.Double immunostaining for PD-1 (blue) and PD-L1 (red) in bright-field shows direct contacts between PD-1+ and PD-L1+ cells in the inflamed meninges (A) and in a WM perivascular infiltrate (B, C; MS234).Double immunostaining for CD20 (brown) and PD-1 (blue) shows that B cells are in contact with PD-1+ cells in B cell follicles from two MS cases (MS92, MS330) (D, F; in the inset in F, the cells indicated by the arrow in F are shown at higher magnification) and in a diffuse meningeal infiltrate (E and insets; MS330).Double immunostaining for LMP2A (brown) and PD-1 (blue) reveals intimate contacts between EBV infected LMP2A+ cells and PD-1+ cells (G, H).Direct cellular contacts were also visualized using double immunofluorescence staining for LMP2A (green) and PD-1 (red) (I, J).Arrows point to cell contacts.Nuclei were stained with DAPI in I and J. Bars: 20 μm.in D-F; 10 μm in A-C, G-J and insets.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)